中文摘要：

该文提出了一种基于语义规则的多民族人脸特征表达方法.该方法在公理化模糊集理论框架下建立刻画人脸特征的语义概念,构建描绘不同民族人脸特征的语义规则集,并根据语义贴近度及隶属度的关联性设定约简准则进行语义挖掘,进而研究不同民族的人脸典型特征.所提方法的特点包括：通过多民族人脸特征数据的分布规律获取表述人脸特征的语义;定义的逻辑运算规则能够实现人脸语义运算,并获得描述人脸多样性特征的复杂语义;挖掘所获的多民族人脸典型特征由具有自然语言解释的语义表述,便于直观理解.文中采用C4.5、Quest、DecisionTable、NeuralNet、BayesNet、SVM和RA等方法对FEI、CK＋以及本文所构建的＂中华多民族人脸数据库＂中的人脸民族属性进行了分析,结果表明该文方法建立的语义规则集不仅能够对各民族人脸特征进行语义解释,而且对个体人脸的民族属性具有较好的判别率,该文方法为研究多民族的人脸语义特征规律提供了一个新途径.

英文摘要：

This paper proposes a semantic rule-based method to represent multi-ethnic facial features.In the frame of axiomatic fuzzy set theory,it builds semantic concepts to describe facial features;establishes semantic rule to express the multi-ethnic facial features;defines a criterion using semantic correlation of close degree and membership degree to mine the semantic rule;then researches the typical multi-ethnic facial features.The advantages method includes that facial semantics are obtained by the distribution of feature data;the complex semantic of facial features can be gained by operating facial semantic based on defined logical rules;the multi-ethnic facial typical features expressed by natural linguistics are easily understandable.Meanwhile,there is an analysis to facial nationality attribute based on face database of FEI,CK＋ and Chinese Multi-Ethnic Face（CMEF）,using C4.5,Quest,DecisionTable,NeuralNet,BayesNet,SVM and RA.The result shows that the semantic rule sets has better semantic interpretability for multi-ethnic facial feature and accuracy for individual recognition.Specially,it provides a new approach to explore regularity of multi-ethnic facial feature.